There's a lot of good reasons for the huge swings. Just due to the nature of the game (puck luck, low scoring games), the vast majority of hockey games are 50-50, or close to it - in that way, hockey's more like soccer than football, basketball, or even baseball. This explains why, fairly often, you'll feel like the Bruins played like crap but still won, or played their ***** off and still lost.
In fact, you can literally turn a 50/50 mode on in the model, and it only changes the percentages a little bit. Outside of bottom dwelling teams like Colorado and Arizona, the "favorite" team is rarely much better than 67% to win.
So, we're literally flipping coins each game. Since we need to win a little over half the points to have a 50-50 chance of making the playoffs, every flip will cause a fairly big swing, especially this late in the season.
Basically, the model is just telling us what we already know from looking at the standings, and from knowing that good teams win slightly more than average, and bad teams lose slightly more than average. Coin flipping gets us to about 55%, and knowing that the Bruins are a slightly above average team gives us a small 5% boost.
So, yeah, the stats site is moderately useless if you look at the standings. And as the season closes, the swings will get bigger. Two teams tied in the standings and playing game 82 against each other will each see a swing of about 50% - one up, and one down.